Portable simulation framework for diffusion MRI.

Area of Science:

• Medical Imaging
• Computational Science
• Biophysics

Background:

• Numerical simulation of diffusion MRI signals aids in understanding complex tissue microstructure and optimizing MRI sequences.
• Finite element (FE) discretization of the Bloch-Torrey equation offers an alternative to traditional random walk simulations.
• Existing FE software packages like FEniCS present implementation challenges for diffusion MRI simulation.

Purpose of the Study:

• To address difficulties in applying FEniCS for diffusion MRI simulations.
• To simplify the use of FEniCS for the MRI community.
• To provide a portable and accessible simulation framework.

Main Methods:

• Developed a portable simulation framework using Python and FEniCS containers for simplified installation.
• Integrated the framework with cloud computing resources (Google Colaboratory, Google Cloud Platform).
• Utilized MPI parallelization for enhanced computational performance.

Main Results:

• Demonstrated simplified software installation via portable FEniCS containers.
• Showcased a portable, open-source Python simulation framework.
• Presented examples validating accuracy, computational efficiency, and parallel computing capabilities.

Conclusions:

• The developed framework facilitates reproducible science in computational diffusion MRI.
• Open-source accessibility and cloud integration accelerate method development and foster research collaborations.
• The framework bridges the gap between MRI simulation needs and available FE software tools.